TPS92640 Current overshoot on first switching cycle.

Is there any way you could provide a schematic? I would also like to ask what you mean about signal ground isolated with a jumper? What is the PWM frequency and duty cycle you are seeing this at? Thanks.

How do i include an attachment?

Here is the schematic. Notice that I'm using a current sense amp to move the current sense from high side to low side. This is because the customer's existing LED fixture uses a common ground for all of the LED strings, so I don't have a separate GND wire for each string.

R182 is my connection from "quiet" ground to power ground.

Also, for now, C61 is removed and R74 is shorted.

I've also attached a scope shot:

Cyan = inductor current (0.5A per division)

Red = output of the current sense amp;

Black = switch node.

you can see that the first pulse is about 2.4usec long, causing inductor current to go up to 2.8A. After that, it discharges, then resumes normal switching around 1A average.

The current sense amp may be contributing, but you can see that it responds very fast, it lags a little behind the actual inductor currrent, but it overshoots way past where the current should regulate. Once the current sense signal passes the set point, it is about 800nsec before that first pulse terminates.

Thanks -

I can't say for sure if the current sense IC has anything to do with the issue or not, I am not familiar with it. But I do have some questions and concerns:

1. How long is the LED current off for before you pull UDIM high again? If it is longer than 8ms you could be going into low power shutdown mode. I ask this because the low side switch turns on briefly causing the inductor current to go below zero. It could be that the boot cap needs to be recharged and that is doing something strange. I also noticed that the output voltage is very low by the time switching starts again.

2. I can't see the layout, but I'm a little concerned about using R182 to tie the grounds together. I generally recommend a sold ground plane for everything. But my main concern in this case is that the CS ground is on the signal ground side. Its ground needs to be tied closely to power ground, particularly near the DAP connection and input capacitor grounds if possible. This obviously has a large voltage slew along with the current rising and that could easily cause an error between the power ground and signal grounds for a brief time and cause an overshoot until things even out.

You should be able to solder thick wires to shore up the grounding and see if that gives you any improvement. It also wouldn't hurt to see if it works fine without the current sense IC. But on that note is there a reason you didn't consider the LM3409 for this design? It already has high side current sense so no external sensing IC would be necessary. It isn't synchronous, but at such a high output voltage relative to the input it would be about the same efficiency anyway.

I was not involved with the development so I'm not as familiar with it as I could be, but I do know who tested it and I believe they would have caught a startup issue. Perhaps it has something to do with the external current sensing? Or perhaps you hit a sweet spot where VCC is near UVLO and the input voltage dip at turn on drops it briefly into UVLO and causes temporary loss of control? I am not sure, but I know I have not seen in on the bench in my limited time with the device.

As for the OVP, it sounds like it could be a noise issue since there is not much hysteresis. Try adding a capacitor from VOUT to ground somewhere in the range of 0.01uF and 0.1uF. I would start with the lower value and increase until there is improvement. Of course it could be a layout issue as well, but we can look at that if capacitance doesn't help. But the output will not exceed the input so if you can tolerate that high of a voltage (basically if your output caps are rated high enough) then you can just disable OVP by grounding it and save the resistors.